NewsletterBlogGlossary

Agentic

What is agentic AI? Systems that autonomously plan, decide, and act to complete multi-step tasks.

concepts
ShareXLinkedIn

Agentic — AI Glossary

Agentic describes AI systems that autonomously plan, make decisions, and take actions to accomplish goals across multiple steps — without requiring human input at each stage. Rather than responding to a single prompt with a single output, agentic AI breaks down complex tasks, uses tools, adapts to intermediate results, and drives workflows toward completion.

Why Agentic Matters

The shift from reactive AI (prompt in, text out) to agentic AI fundamentally changes what software can do unsupervised. Instead of generating a code snippet for you to paste, an agentic system like Claude Code reads your codebase, plans a refactoring strategy, edits files, runs tests, and commits the result.

This matters for productivity and for the architecture of AI applications. Agentic workflows enable AI to handle multi-step processes — research, customer support, software engineering, data analysis — that previously required a human in the loop at every decision point. Anthropic's recent work on desktop agents and cowork mode shows where the industry is heading: AI that operates alongside you, not just underneath your cursor.

How Agentic Works

Agentic behavior emerges from combining a large language model with three capabilities:

  • Planning: The model decomposes a high-level goal into ordered subtasks, reassessing the plan as new information arrives
  • Tool use: The model calls external tools — shell commands, APIs, file systems, browsers — to interact with the real world rather than just generating text
  • Memory and state: The model tracks what it has done, what worked, and what remains, maintaining context across a long sequence of actions

Claude powers several agentic implementations, including Claude Code for software engineering and the Claude desktop agent for general computer use. The key architectural pattern is a loop: the model observes its environment, decides on an action, executes it, observes the result, and repeats until the task is complete.

  • Claude Code: Anthropic's terminal-based agentic coding tool that plans and executes multi-file engineering tasks
  • Claude: Anthropic's family of large language models that powers agentic workflows through extended context and tool use
  • Anthropic: The AI safety company building Claude and pioneering agentic AI systems

Want more AI insights? Subscribe to LoreAI for daily briefings.